Three-Dimensional Imaging of Metabolites inTissues under Ambient Conditions by LaserAblation Electrospray Ionization MassSpectrometry

Peter Nemes, Alexis A. Barton, and Akos Vertes*

Department of Chemistry, W. M. Keck Institute for Proteomics Technology and Applications,George Washington University, Washington, DC 20052

Three-dimensional (3D) imaging of molecular distribu-tions offers insight into the correlation between biochemi-cal processes and the spatial organization of a biologicaltissue. Simultaneous identification of diverse moleculesis a virtue of mass spectrometry (MS) that in combinationwith ambient ion sources enables the atmospheric pres-sure investigation of biomolecular distributions and pro-cesses. Here, we report on the development of an MS-based technique that allows 3D chemical imaging oftissues under ambient conditions without sample prepa-ration. The method utilizes laser ablation electrosprayionization (LAESI) for direct molecular imaging withlateral and depth resolutions of ∼300 µm and 30-40 µm,respectively. We demonstrate the feasibility of LAESI 3Dimaging MS of metabolites in the leaf tissues of Peace lily(Spathiphyllum lynise) and the variegated Zebra plant(Aphelandra squarrosa). Extensive tandem MS studieshelp with the structure identification of the metabolites.The 3D distributions are found to exhibit tissue-specificmetabolite accumulation patterns that correlate with thebiochemical roles of these chemical species in plantdefense and photosynthesis. Spatial correlation coef-ficients between the intensity distributions of different ionshelp to identify colocalization of metabolites and poten-tially uncover connections between metabolic pathways.

Many of the current three-dimensional (3D) biomedical imag-ing methods achieve exquisite lateral and depth resolution for invivo imaging by relying on the interaction of electromagneticradiation or particles with the specimen.1 For example, coherentanti-Stokes Raman scattering has demonstrated capabilities tofollow lipid distributions on the cellular or subcellular level.2

Several of these techniques, however, exhibit limited chemicalselectivity, are specific to a small number of molecular species,and often require the introduction of fluorescent or other molec-ular labels. These restrictions are less prevalent in methods basedon mass spectrometry (MS) that simultaneously report the

distributions for diverse molecular species without the need forfluorescent or other labels.3,4

Established imaging MS methods, such as secondary ion massspectrometry (SIMS) and matrix-assisted laser desorption ioniza-tion (MALDI), and recent developments including nanostructureinitiator mass spectrometry have demonstrated exceptional capa-bilities in capturing the 2D and 3D distributions of endogenousand drug molecules in tissue and whole-body sections.5-10

Recently, using C60 primary ions, time-of-flight SIMS has enabled3D studies of a freeze-dried oocyte and revealed the distributionof lipids and related molecules with excellent spatial resolution.4

Imaging vacuum MALDI MS, in combination with cryosec-tioning, can produce a series of 2D protein and peptide imagesthat can be reconstructed in 3D.9 Characteristic to thesemethods, however, is the requirement for sample preparationand the need to perform the experiment in vacuum, preventingthe study of live specimens.

Atmospheric pressure imaging MS circumvents these limita-tions by bringing the ionization step into the ambient environmentwhile minimizing or eliminating treatment to the sample.11 Duringthe past few years, this field has experienced rapid developmentproviding several ambient ion sources for imaging MS. Forexample, atmospheric pressure (AP) infrared MALDI12-14 hascaptured metabolite transport in plant vasculature,14 desorption

* To whom correspondence should be addressed. E-mail: vertes@gwu.edu.Phone: 202-994-2717. Fax: 202-994-5873.

(1) Weissleder, R.; Pittet, M. J. Nature 2008, 452, 580–589.(2) Evans, C. L.; Potma, E. O.; Puoris’haag, M.; Cote, D.; Lin, C. P.; Xie, X. S.

Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 16807–16812.

(3) Northen, T. R.; Yanes, O.; Northen, M. T.; Marrinucci, D.; Uritboonthai,W.; Apon, J.; Golledge, S. L.; Nordstrom, A.; Siuzdak, G. Nature 2007,449, 1033–U1033.

(4) Fletcher, J. S.; Lockyer, N. P.; Vaidyanathan, S.; Vickerman, J. C. Anal.Chem. 2007, 79, 2199–2206.

(5) Ostrowski, S. G.; Van Bell, C. T.; Winograd, N.; Ewing, A. G. Science 2004,305, 71–73.

(6) Khatib-Shahidi, S.; Andersson, M.; Herman, J. L.; Gillespie, T. A.; Caprioli,R. M. Anal. Chem. 2006, 78, 6448–6456.

(7) McDonnell, L. A.; Heeren, R. M. A. Mass Spectrom. Rev. 2007, 26, 606–643.

(8) Zhang, H.; Cha, S. W.; Yeung, E. S. Anal. Chem. 2007, 79, 6575–6584.(9) Andersson, M.; Groseclose, M. R.; Deutch, A. Y.; Caprioli, R. M. Nat.

Methods 2008, 5, 101–108.(10) Yanes, O.; Woo, H.-K.; Northen, T. R.; Oppenheimer, S. R.; Shriver, L.; Apon,

J.; Estrada, M. N.; Potchoiba, M. J.; Steenwyk, R.; Manchester, M.; Siuzdak,G. Anal. Chem. 2009, 81, 2969–2975.

(11) Cooks, R. G.; Ouyang, Z.; Takats, Z.; Wiseman, J. M. Science 2006, 311,1566–1570.

(12) Laiko, V. V.; Taranenko, N. I.; Berkout, V. D.; Yakshin, M. A.; Prasad, C. R.;Lee, H. S.; Doroshenko, V. M. J. Am. Soc. Mass Spectrom. 2002, 13, 354–361.

(13) Li, Y.; Shrestha, B.; Vertes, A. Anal. Chem. 2008, 80, 407–420.(14) Li, Y.; Shrestha, B.; Vertes, A. Anal. Chem. 2007, 79, 523–532.

Anal. Chem. 2009, 81, 6668–6675

10.1021/ac900745e CCC: $40.75 2009 American Chemical Society6668 Analytical Chemistry, Vol. 81, No. 16, August 15, 2009Published on Web 07/02/2009

Dow

nloa

ded

by G

EO

RG

E W

ASH

ING

TO

N U

NIV

on

Aug

ust 1

8, 2

009

Publ

ishe

d on

Jul

y 2,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

ac90

0745

e

electrospray ionization (DESI)15 has imaged drugs of abusein fingerprints16 and metabolite distributions in thin tissue sec-tions,17-19 and laser ablation electrospray ionization (LAESI)20,21

has provided insight into metabolic differences between thesectors of variegated plants and described a simple form of depthprofiling. Lateral imaging methods alone do not capture thelocalization of chemical species in the depth dimension.

Our previous paper discussed depth profiling at a particularposition and lateral imaging separately.21 This raised the prospectof combining these two modes into 3D imaging. This proposition,however, has very important technical consequences regardingsource stability, as it is much more challenging to maintain sourcestability for several hours required for the latter. There are alsomajor differences in the potential amount and quality of informa-tion gathered. A 3D data set might exhibit correlations amongmetabolites and in relation to morphology that a simple depthprofile does not capture. Improvements in ionization efficiency,signal stability, and depth resolution are required to provide repro-ducible quality depth profiling over at least 500 positions for thedevelopment of 3D molecular imaging with the LAESI ion source.

Here we report the first example of 3D imaging with MS underatmospheric pressure conditions that illuminates the relationshipbetween tissue architecture and metabolism. In these experiments,LAESI 3D imaging MS is realized as a combination of lateralimaging and depth profiling by tissue ablation with single laserpulses of ∼2.94 µm wavelength. The O-H vibrations of the nativewater molecules in the tissue samples readily absorbed the laserpulse energy leading to ablation, i.e., the ejection of a microscopicvolume of the sample in the form of neutral particulates and/ormolecules.22 This plume was then intercepted by an electrospray,and the ablated material was efficiently postionized. Proof-of-principle experiments on leaves marked with ink confirmed thatby gradually ablating through the tissue at every location, LAESI3D imaging MS was able to recover the patterns on both the upper(adaxial) and the lower (abaxial) surfaces. Tandem MS wasperformed on numerous ions to help with structure assignments.The acquired 3D distributions of endogenous metabolites wereinterpreted in the context of metabolic pathways and theircorrelation with tissue morphology.

EXPERIMENTAL SECTIONLaser Ablation Electrospray Ionization Mass Spectrom-

etry. The LAESI source was identical to the one we have recentlydescribed.20,21 The leaf tissues of peace lily (Spathiphyllum lynise)and zebra plant (Aphelandra squarrosa) were mounted on a three-axis translation stage, positioned 18 mm below the electrospray

axis. A Nd:YAG laser driven optical parametric oscillator (VibrantIR, Opotek Inc., Carlsbad, CA) provided mid-infrared pulses at2940 nm wavelength and 0.2 Hz repetition rate. This laser beamwas used to ablate samples at 90° incidence angle, ∼3-5 mmdownstream from the tip of the spray emitter. During the S. lynise(∼200 µm average leaf thickness) and A. squarrosa (∼450 µmaverage leaf thickness) imaging experiments, the average outputenergy of a laser pulse was 0.1 mJ ± 15% and 1.2 mJ ± 10%, whichtranslated into a fluence of ∼0.1 and ∼1.2 J/cm2 at the focal point,respectively. Optical microscopy of the ablation marks inZ-stack mode confirmed that these single laser pulses removed∼30-40 µm of the tissue (see the animated GIF image in theSupporting Information). The ablated material was interceptedby the electrospray plume, and the resulting ions were analyzedby an orthogonal acceleration time-of-flight mass spectrometer (Q-TOF Premier, Waters Co., Milford, MA) with a 1 s/spectrumintegration time. The sampling cone of the mass spectrometerwas located on axis with and 13 mm away from the tip of thespray emitter. The mass spectra were externally calibrated withsodium iodide clusters with a mass accuracy of ∼5 mDa and aresolution of ∼6000 (fwhm) between m/z 100 and 1000 in allexperiments. Assignments of the detected ions to particularmetabolites was facilitated by accurate mass measurements,comparisons of isotope distribution patterns, and tandem MS.Fragmentation was induced by collision-activated dissociation(CAD) in argon collision gas at 4 × 10-3 mbar pressure with thecollision energy set between 15 and 30 eV.

Three-Dimensional Molecular Imaging with LAESI MS.A three-axis translation stage was positioned with motorizedactuators (LTA-HS, Newport Corp., Irvine, CA) to scan the samplesurface on a 2D grid. The actuators had a travel range of 50 mmand a minimum incremental motion of 0.1 µm. Thus, the ultimateresolution was determined by the focusing of the incident laserbeam and the dimensions of the ablation craters (∼300 µm inaverage diameter). At each lateral grid point, a depth profile wascreated with six laser pulses, while the generated ions wererecorded with the mass spectrometer. Under these conditions,3D imaging of a 12.5 × 10.5 × 0.45 mm3 volume required a totalanalysis time of ∼5 h. Software was written in-house (LabView8.0) to position the translation stage and render the analysistimes to the corresponding X-Y-Z coordinates. The exporteddata sets of mass-selected ions were presented as layer-by-layercontour plot images with a scientific visualization package(Origin 7.0, OriginLab Co., Northampton, MA). Because thetissue composition was probed with 300 µm pixels at every 500µm in the X and Y directions, ∼200 µm tissue areas were leftintact during the imaging experiments (see Figure 1A). Theimage reconstruction algorithm applied in the current studyinterpolated the molecular composition over the intact areas.

Chemicals and Plants. Glacial acetic acid (TraceSelect grade)and gradient grade methanol and water were obtained from Sigma-Aldrich and used as received. The S. lynise and A. squarrosa plantswere purchased from a local florist at an approximate age of oneand a half years. The plants were watered every 2 days with ∼300mL of tap water to keep their soil moderately moist to the touch.No fertilizer was used during the experiments. Temperature andlight conditions were 20-25 °C in light shade, protected fromdirect sun.

(15) Takats, Z.; Wiseman, J. M.; Gologan, B.; Cooks, R. G. Science 2004, 306,471–473.

(16) Ifa, D. R.; Manicke, N. E.; Dill, A. L.; Cooks, G. Science 2008, 321, 805–805.

(17) Wiseman, J. M.; Ifa, D. R.; Cooks, R. G.; Venter, A. Nat. Protoc. 2008, 3,517–524.

(18) Kertesz, V.; Van Berkel, G. J.; Vavrek, M.; Koeplinger, K. A.; Schneider,B. B.; Covey, T. R. Anal. Chem. 2008, 80, 5168–5177.

(19) Wiseman, J. M.; Ifa, D. R.; Zhu, Y. X.; Kissinger, C. B.; Manicke, N. E.;Kissinger, P. T.; Cooks, R. G. Proc. Natl. Acad. Sci. U.S.A. 2008, 105,18120–18125.

(20) Nemes, P.; Vertes, A. Anal. Chem. 2007, 79, 8098–8106.(21) Nemes, P.; Barton, A. A.; Li, Y.; Vertes, A. Anal. Chem. 2008, 80, 4575–

4582.(22) Apitz, I.; Vogel, A. Appl. Phys. A: Mater. Sci. Process. 2005, 81, 329–338.

6669Analytical Chemistry, Vol. 81, No. 16, August 15, 2009

Dow

nloa

ded

by G

EO

RG

E W

ASH

ING

TO

N U

NIV

on

Aug

ust 1

8, 2

009

Publ

ishe

d on

Jul

y 2,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

ac90

0745

e

RESULTS AND DISCUSSIONThree-Dimensional Molecular Imaging by LAESI. Initially,

the 3D molecular imaging capabilities of LAESI were evaluated inproof-of-principle experiments. As shown in Figure 1A, the adaxialand abaxial surfaces of an S. lynise leaf were marked with ∼1 mmwide right-angle lines and a 4 mm diameter spot with basic blue 7

and rhodamine 6G dyes, respectively. Laser pulses of 2.94 µmwavelength were focused on the adaxial (upper) surface of thissample, and the laser energy was adjusted to acquire a six-step depthprofile for each point on a 22 × 26 grid across a 10.5 × 12.5 mm2

(23) Hetherington, A. M.; Woodward, F. I. Nature 2003, 424, 901–908.

Figure 1. LAESI 3D imaging MS of S. lynise leaf tissue in proof-of-principle experiments. The adaxial and the abaxial cuticles were markedwith right-angle lines and a spot colored in basic blue 7 and rhodamine 6G, respectively. (A) Top view of the interrogated area with an array ofablation marks. Some rhodamine 6G dye from the bottom is visible through the ablation holes. (B) Ion distributions for basic blue 7 (m/z 478.3shown in blue color) and rhodamine 6G (m/z 443.2 shown in the orange/wine color range) on false color scales paralleled the mock patternsshown in the optical image. Higher abundances of the endogenous metabolite leucine (m/z 154.1 shown in gray color scale) were observed inthe top two layers. (C) Cyanidin/kaempferol rhamnoside glucoside (m/z 595.2 shown in gray color) populated the epidermal region. (D) Chlorophylla (m/z 893.5 shown in the cyan/royal blue color range) was abundant in the palisade mesophyll region, in agreement with the known localizationof chloroplasts within plant tissues.

6670 Analytical Chemistry, Vol. 81, No. 16, August 15, 2009

Dow

nloa

ded

by G

EO

RG

E W

ASH

ING

TO

N U

NIV

on

Aug

ust 1

8, 2

009

Publ

ishe

d on

Jul

y 2,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

ac90

0745

e

area. Each of the resulting 3432 cylindrical voxels with 350 µmdiameter and 40 µm height, i.e., ∼4 nL analysis volume, yieldeda high-resolution mass spectrum. The top view of the leaf followingLAESI 3D imaging MS can be seen in Figure 1A. A lateral stepsize of 500 µm yielded ∼2-3 pixels to sample across the width ofthe lines drawn in basic blue 7. A circular rhodamine 6G dye patternfrom the marking of the backside can be seen in the lower left cornerof the image, indicating complete tissue removal.

The lateral and cross-sectional localization of the dye ions wasfollowed in 3D. The color-coded contour plots in Figure 1Bdemonstrate the localization of the dye ions and some endogenousmetabolites in the plant organ. In Figure 1B, each layer representsan ∼40 µm thick section of the leaf tissue sampled by successiveablations. The lateral distribution of the basic blue 7 dye ion,[C33H40N3]+, detected at m/z 478.3260 and shown in blue inthe top layer of Figure 1B was in very good correlation with itsoptical pattern (compare with Figure 1A). Careful inspection ofthe optical microscope images and the acquired spectra indicatedthat the laser ablation did not produce lateral mixing or crosscontamination in the tissue to a significant degree. Although thebasic blue 7 dye was only applied to the top cuticle of the leaf, itsmolecular ion was also noticed at low intensities in the secondlayer (<200 counts). The corresponding pixels of the imageparalleled the high-intensity areas (>7000 counts) of the first layer,which suggested that a small fraction of the dye was transportedwithin the tissue during the mock sample preparation step.

The molecular ion of the rhodamine 6G dye, [C28H31N2O3]+,with a measured m/z 443.2295 and shown in red in Figure 1B,was found at high abundances in the fifth and six layers. Figure1B shows that the lateral distribution patterns of the dye ion inthe bottom two layers agree well with the marked spot on theabaxial cuticle seen through the ablation craters in the opticalimage (see Figure 1A for comparison). Plant leaves are known tohave generally thinner abaxial cuticles with a higher density ofstomata than the upper surface.23 In addition to their natural role toregulate gas and water exchange with the environment, the stomatamight have facilitated cross-sectional transport of the dye solutionto deeper layers of the leaf during the preparation of the mocksample. Furthermore, similar to the top surface of the leaf, limitedcross-sectional mixing due to laser ablation cannot be excluded.

Our cumulative experience suggests that LAESI achieves softionization of molecules from a variety of samples, includingbiological tissues that are similar in nature to those presented foranalysis in this study.20,21,24 Similarly, when the shot-to-shot thestability of ion signal for individual m/z values in LAESI ionizationwas tested in homogeneous samples, consecutive laser pulsesshowed modest fluctuation in ion intensities but no noticeabletrend in the average, thus indicating the lack of laser-induceddegradation (data not shown).

To further prove this point, small organics of know composition(reserpine and verapamil) were deposited on the studied leaftissue. When LAESI spectra were collected and the backgroundfrom the leaf substrate was subtracted from these spectra, sta-ble molecular ion signal was observed with no fragmentation.These results were added to the Supporting Information. Clearly,the validity of this point also depends on the thermal stability of

the molecules of interest and the applied laser fluence. Judiciousselection of the latter and critical data analysis are necessarysafeguards here. For example, the ablation or postionizationprocesses might interfere with the integrity of certain molecularclasses. In particular, large noncovalent complexes might undergodissociation in the charged droplets.

Metabolites and Their Localization. Endogenous metabo-lites in the leaf tissue were assigned based on their accuratemasses, isotope distribution patterns, and in many cases tandemMS. The latter was especially useful in the case of the flavonoidglycosides that exhibit a wide array of structural isomers. Indeed,for these compounds not all of the structural ambiguity could beresolved. Several of the LAESI ions with a signal-to-noise ratio,S/N, >3 were tentatively assigned to particular molecules. Thetentative assignments of some observed metabolites along withthe layers of their accumulation, where appropriate, are sum-marized in Table 1 for the leaves of both S. lynise and A. squarrosa.For the ions with the “e” superscript for the measured m/z values,tandem MS was performed. As an example, Figure 2 shows theCAD spectrum of the m/z 493 ion, tentatively assigned as, e.g.,methoxykaempferol glucuronide or methoxyluteolin glucuronide.Many other positional isomers exist. The presence of the m/z234 fragment in the CAD spectrum assigned as 2,4B0, however,is an indication that the glucuronide is likely in the 3-O positionand the structure is derived from kaempferol.25 The methoxygroup can still be in the 6 or the 8 position on the flavonestructure. The inset in Figure 2 shows the 6-methoxykaempferol-3-glucuronide structure and its fragmentation that is consistentwith the CAD spectrum.

The experiments with the mock 3D dye distributions on theS. lynise leaf demonstrated that LAESI allowed 3D imaging MSwith low spatial resolution in the lateral and cross-sectionaldimensions of the tissue, also revealing the distributions of variousendogenous plant metabolites. The acquired mass spectra revealedcompositional changes as a function of depth (compare massspectra for pulses 1-6 at a particular location in Figure S1 of theSupporting Information). For example, the 3D distribution of theprotonated leucine ion with m/z 154.1 can be seen in Figure 1Bon a gray-to-black false color scale. This amino acid was observedacross the entire tissue (with S/N > 3) with higher ion counts inthe top 80 µm section. In contrast, the molecular ion of cyanidin/kaempferol rhamnoside glucoside with m/z 595.2 along with othersecondary metabolites was uniquely linked to the top 40 µm of thetissue (shown in gray color in Figure 1C). Chlorophyll a with m/z893.5 was characteristic to the palisade mesophyll layer of the leaftissue from 40 to 80 µm (shown in the cyan/royal blue color rangein Figure 1D). This result further confirmed the feasibility of LAESI3D imaging MS in plant leaves as independent studies based onsectioning experiments had shown similar results.26

Independent methods showed that a higher concentration ofkaempferol glycosides is often found in the upper epidermallayers.26-28 In leaves of rapeseed (Brassica napus), for example,quercetin- and kaempferol-based UV-screening pigments areconcentrated within the upper 40 µm of the leaf tissue,26 showinga very good agreement with our data for S. lynise. Plant flavonoidsare thought to play a vital role in providing protection against the

(24) Sripadi, P.; Nazarian, J.; Hathout, Y.; Hoffman, E.; Vertes, A. Metabolomics2009, 5, 263–276.

(25) March, R. E.; Lewars, E. G.; Stadey, C. J.; Miao, X. S.; Zhao, X. M.; Metcalfe,C. D. Int. J. Mass Spectrom. 2006, 248, 61–85.

6671Analytical Chemistry, Vol. 81, No. 16, August 15, 2009

Dow

nloa

ded

by G

EO

RG

E W

ASH

ING

TO

N U

NIV

on

Aug

ust 1

8, 2

009

Publ

ishe

d on

Jul

y 2,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

ac90

0745

e

detrimental effects of solar radiation. By direct light absorptionor scavenging harmful radicals these substances can create abarrier against the effect of UV-A and -B rays, protecting thephotosynthetic mesophyll cells and perhaps providing them withadditional visible light via fluorescence.28,29

Metabolism and Tissue Architecture. We used LAESI 3Dimaging MS to explore spatial variations in the metabolites of thevariegated leaves of the A. squarrosa plant. The optical image ofthe adaxial side of the leaf showed alternating sectors of greenand yellow color (see Figure 3A). The chemical makeup of an11.5 × 7.5 mm2 area of this tissue was probed in 3D on a 24 ×16 × 6 grid, resulting in 2304 voxels. Figure 3B presents theinterrogated area, marked by an array of ablation craters. Asconsecutive single laser pulses sampled the composition of thetissue, mass spectra were simultaneously recorded. The detected

ions with S/N > 3 were tentatively assigned to particular moleculesbased on accurate mass measurements, isotope distributionanalysis, and CAD experiments combined with broad plantmetabolomic database searches at http://www.arabidopsis.org,http://biocyc.org, and http://www.metabolome.jp Web sites (lastaccessed on April 4, 2009).

Some of the metabolites accumulated in the mesophyll layersof the leaf tissue. In this region mass analysis showed the presenceof various ions in the m/z 600-1000 portion of the spectrumincluding m/z 650.4, 813.5, 893.5, and 928.6. On the basis of theaccurate mass (see Table 1) and the isotopic distribution patternof the m/z 893.5425 ion (76% ± 4% and 50% ± 8% for M + 1 and M+ 2, respectively) we assigned it as the protonated chlorophyll amolecule (C55H73N4O5Mg+ with predicted 77% and 43% for M+ 1 and M + 2, respectively). CAD of the m/z 893.5 ion yieldedan abundant fragment at m/z 615.2, corresponding to theprotonated form of the chlorophyllide a, C35H35N4O5Mg+, asdocumented by other researchers.30-32 The 3D distribution ofthe chlorophyll a ion showed an accumulation of this species in

(26) Alenius, C. M.; Vogelmann, T. C.; Bornman, J. F. New Phytol. 1995, 131,297–302.

(27) Strack, D.; Pieroth, M.; Scharf, H.; Sharma, V. Planta 1985, 164, 507–511.

(28) Agati, G.; Galardi, C.; Gravano, E.; Romani, A.; Tattini, M. Photochem.Photobiol. 2002, 76, 350–360.

(29) Hoque, E.; Remus, G. Photochem. Photobiol. 1999, 69, 177–192.(30) Koichi, T.; Takakazu, K.; Masanori, M. Anal. Sci. 1987, 3, 527–530.(31) Schoefs, B. J. Chromatogr., A 2004, 1054, 217–226.

Table 1. Assignments of Selected Metabolites and Their Tissue-Specific Accumulation in S. lynise and A.squarrosaa

metabolite formula monoisotopic m/z measured m/z ∆m (mDa) accumulationb

γ-aminobutiric acidc C4H9NO2 104.0712 (H) 104.0701 -1.1 ADE, PMcoumarin C9H6O2 147.0446 (H) 147.0378 -6.8 ADEketoglutaric acidc C5H6O5 147.0293 (H) 8.5leucine, isoleucinec C6H13NO2 154.0844 (Na) 154.0819 -2.5 ADE, PM7-oxocoumarind C9H6O3 163.0395 (H) 163.0391 -0.4 uniformxylose, ribosec C5H10O5 189.0165 (K) 189.0090 -7.5 N/Aglucosec C6H12O6 203.0532 (Na) 203.0526 -0.6 ADE

219.0271 (K) 219.0274 0.3ribulose phosphatec C5H11O8P 268.9829 (K) 268.9868 3.9 uniformacacetind C16H12O5 285.0763 (H) 285.0756e -0.7 PM, SM, and in yellow,

uni-ADE, uni-ABEkaempferol, luteolin C15H10O6 287.0556 (H) 287.0565e 0.9 PM, SM in yellow sectorscyanidind C15H11O6 287.0556 (+)methylkaempferol, methylluteolin, flavones C16H12O6 301.0712 (H) 301.0721e 0.9 PM and in yellowpeonidind C16H13O6 301.0712 (+)methoxy-kaempferol/luteolin C16H12O7 317.0661 (H) 317.0741e 8.0 PM, SM, and in yellowpetunidind C16H13O7 317.0661 (+)methoxy-hydroxyphenyl glucosided C13H18O8 325.0899 (Na) 325.0886e -1.3 uniformmalvidin C17H15O7 331.0818 (+) 331.0747 -7.1 uniformtrihydroxy-dimethoxyflavoned C17H15O7 331.0818 (H)sucrosec C12H22O11 365.1060 (Na) 365.1058 -0.2 ADE, PM, SM

381.0799 (K) 381.0794 -0.5tetrahydroxy trimethoxyflavoned C18H16O9 377.0873 (H) 377.0791 -8.2 PMflavonoidsc C21H18O7 383.1131 (H) 383.1130 -0.1 ADEpentahydroxy-trimethoxyflavonec C18H16O10 393.0822 (H) 393.0839 1.7 N/Acyanidin rhamnoside, luteolinidin glucosidec C21H21O10 433.1135 (+) 433.1125e -1.0 ADEkaempferol glucuronide, luteolin glucuronided C21H18O12 463.0877 (H) 463.0876e -0.1 PM in yellowmethoxykaempferol glucuronided C22H20O13 493.0982 (H) 493.0955e -2.7 PM, SM, and in yellowkaempferide glucuronide, luteolin methylglucuronidec C22H20O12 499.0853 (Na) 499.0936 8.3 ADEkaempferol rhamnoside glucosidec C27H30O15 595.1663 (H) 595.1649e -1.4 ADE

617.1483 (Na) 617.1472 -1.1cyanidin rhamnoside glucosidec C27H31O15 595.1663 (+) 595.1649e -1.4cyanidin diglucoside C27H31O16 611.1612 (+) 611.1575 -3.7 N/Akaempferol diglucosidec C27H30O16 611.1612 (H)quercetin coumarylglucosidec C30H26O14 633.1221 (Na) 633.1218e -0.3 N/Akaempferol diglucuronide, luteolin diglucuronided C27H26O18 639.1198 (H) 639.1212e 1.4 PM, SM in yellow,

uni-ADE, uni-ABEluteolin methyl ether glucuronosyl glucuronided C28H28O18 653.1354 (H) 653.1339e -1.5 PM, SM in yellow, uni-ADEkaempferol-(diacetyl coumarylrhamnoside)d C34H30O14 663.1714 (H) 663.1641e -7.3 SMchlorophyll ac,d C55H72N4O5Mg 893.5430 (H) 893.5425c,e -0.5c PM, SMc

893.5376d,e -5.4d PM, SM in green,uni-ADE, uni-ABEd

a Spatial cross-correlations are available for species in bold in Table 2. b Tissue layers are noted as adaxial epidermis (ADE), palisade mesophyll(PM), spongy mesophyll (SM), abaxial epidermis (ABE). Not available, N/A, indicates S/N < 3 for some voxels of the image. c Metabolites foundin the leaf tissue of S. lynise. d Metabolites observed in the leaf tissue of A. squarrosa. e Tandem MS experiments facilitated metabolite assignment.

6672 Analytical Chemistry, Vol. 81, No. 16, August 15, 2009

Dow

nloa

ded

by G

EO

RG

E W

ASH

ING

TO

N U

NIV

on

Aug

ust 1

8, 2

009

Publ

ishe

d on

Jul

y 2,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

ac90

0745

e

the second, and to some degree, in the third layers, i.e., this ionwas found between 40 and 120 µm below the adaxial cuticle (seeFigure 3C in blue color). Similar to our observations for S. lyniseabove, this 3D profile reflects the known biological localizationof chlorophyll a in the chloroplasts of the palisade and spongymesophyll layers where photosynthesis takes place.

Visual inspection of the 3D distributions provided informationon the accumulation patterns of the different metabolites withrespect to the morphological features of the tissue. For example,for A. squarrosa Figure 3C showed that kaempferol/luteolin(yellow/orange scale) exhibited heterogeneity both laterally andas a function of depth and was most abundant in the second andthird layers. The blue false color scale in Figure 3C indicated thatchlorophyll a populated the mesophyll layers. For this ion, lowerintensities were observed in the yellow sectors, in agreement withtheir achlorophyllous nature. Figure 3D showed that acacetin,buried in layers two and three, followed the variegation patternbut did not exhibit heterogeneity in the top and bottom epidermallayers. According to Figure 3E that kaempferol-(diacetyl cou-marylrhamnoside), with m/z 663.2, accumulated in the mesophylllayers (third and fourth) with uniform lateral distributions.

Correlations between the tissue structure discerned from theoptical images and the 3D molecular images of certain secondarymetabolites indicated that the biosynthesis of these species waslinked to the yellow sectors of the plant. For example, kaempferol/luteolin was exclusively found in the achlorophyllous tissue withhigher abundances in the adaxial epidermis and the palisade andspongy mesophyll layers (see Figure 3C). These results suggestedthat the presence of kaempferol/luteolin was likely the result ofmetabolic processes active in the yellow sectors.

Figure 3D showed that the acacetin distribution was ratheruniform in the first, fourth, fifth, and sixth layers. The secondand third layers, however, exhibited lateral heterogeneity in themolecular distribution. The high-intensity pixels (see intensitiesabove ∼200 counts in red color) resembled the variegation pattern.The secondary metabolites kaempferol/luteolin diglucuronide andluteolin methyl ether glucuronosyl glucuronide exhibited similardistributions in space. These features were only partially revealedor completely obscured in our earlier 2D imaging experimentsthat integrated for the cross section of the tissue.21

The protonated ions of the methyl, methoxy, and glucuronidederivatives of kaempferol/luteolin showed similar distributionpatterns (see Supporting Information, Figure S2). The opticalimage of the leaf cross section (data not shown) revealed thatthe secondary vasculature was located ∼150-200 µm below theupper surface and was in direct contact with the cells of the yellowsectors. The correlation between the molecular and the opticalimages suggested that the glucuronide derivative localized closeto the secondary veins of the leaf.

In Figure 3E, the 3D molecular image of kaempferol-(diacetylcoumarylrhamnoside) revealed significantly higher ion counts andhomogeneous presence in the spongy mesophyll (third andfourth) layers compared to the epidermal sections. For tetrahy-droxy-trimethoxyflavone, the center of distribution, however,shifted toward the palisade mesophyll (second and third layers).A few ions, including those registered at m/z 501 and 647, alsobelonged to this group (see Supporting Information, Figures S3and S4, respectively).

Although due to matrix and suppression effects the ionintensities often do not directly represent concentrations, relativeconcentration values and their correlations throughout the tissuesmight be evaluated. Quantitative characterization of the relation-ship between metabolite distributions is possible through thecorrelation between the normalized intensity distributions for theith and jth ions obtained by LAESI 3D imaging MS, Ii(r) andIj(r), respectively. The Pearson product-moment correlationcoefficient is defined as

Fi,j )Cov(Ii(r), Ij(r))

σIiσIj

where Cov is the covariance of the two variables in the imagedvolume and σIi and σIj stand for the standard deviations of Ii(r)and Ij(r), respectively. Fi,j is a measure of the connectionbetween the distributions of the two metabolites. Spatialcorrelations between the intensity distributions of ith and jthions, Fi,j, can help to identify the metabolic relationship betweenthe two chemical species.

For the metabolites in boldface in Table 1, the correlationbetween their 3D localization in the A. squarrosa leaf was evaluatedthrough Pearson product-moment correlation coefficients. Table2 reports the correlation coefficients calculated for the 12 selectedm/z. Table S1 in the Supporting Information lists the correlationcoefficients for a larger set of 26 ions.

For obvious cases, e.g., m/z 301 and 317 with F301,317 ) 0.88,visual comparison of the distributions confirmed the strongcorrelation between ion distributions. The degree of similaritywas also followed for less clear cases. For example, for m/z285 and 287 with F285,287 ) 0.65, both distributions reflected

(32) Vavilin, D.; Brune, D. C.; Vermaas, W. Biochim. Biophys. Acta 2005, 1708,91–101.

Figure 2. Tandem MS facilitated the assignment of metabolite ionsin the LAESI experiments. For example, collision-activated dissocia-tion of the m/z 493 ion yielded fragments consistent with the structureof methoxykaempferol-glucuronide. The inset shows the majorfragmentation pathways proposed for 6-methoxykaempferol-3-glu-curonide. The nomenclature for flavonoid glycoside fragment ions wasadopted from March et al. (ref 25). The correspondence between thespectrum and the fragment ions is as follows: m/z 133.0 is interpretedas [0,2A1 - H]+, m/z 163.0 is interpreted as [0,1A1 - H]+, m/z 177.1 isinterpreted as B1

+, m/z 223.1 is interpreted as [Y0 - C6H5O]+, m/z234.1 is interpreted as [2,4B0 + H]+, m/z 315.1 is interpreted as Y0

+,m/z 317.1 is interpreted as [Y0 + 2H]+, m/z 331.1 is interpreted as[0,1X0 + 3H]+, and m/z 345.1 is interpreted as [1,5X0 + H]+.

6673Analytical Chemistry, Vol. 81, No. 16, August 15, 2009

Dow

nloa

ded

by G

EO

RG

E W

ASH

ING

TO

N U

NIV

on

Aug

ust 1

8, 2

009

Publ

ishe

d on

Jul

y 2,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

ac90

0745

e

the variegation pattern, but in layers two and three the m/z285 distribution exhibited significant values in the greensectors, as well. Another interesting example was the lack ofspatial correlation between kaempferol/luteolin at m/z 287 andchlorophyll a at m/z 893. The low value of the correlationcoefficient, F287,893 ) 0.08, indicated that these two metaboliteswere not colocalized. They are also known to belong to differentmetabolic pathways. Conversely, a high correlation coefficient,

F287,301 ) 0.94, between kaempferol and methylkaempferolsignified their colocalization and possibly their interconversionthrough the kaempferol 4′-O-methyltransferase enzyme.33

These and other examples (see them in the SupportingInformation) showed that the correlation coefficients couldbecome a tool to uncover the connections between metabolicpathways through measuring the colocalization of metabolites intissues.

Figure 3. Metabolites in relation to tissue architecture captured by LAESI 3D imaging MS. Optical image of A. squarrosa leaves (A) before and(B) after analysis. (C) LAESI 3D imaging MS distribution of kaempferol/luteolin with m/z 287.0 (yellow/orange scale) followed the variegationpattern. Chlorophyll a with m/z 893.5 (blue scale) accumulated in the mesophyll layers. (D) Acacetin with m/z 285.0 showed higher abundancein the yellow sectors of the second and third layers with a homogeneous distribution in the others. (E) Kaempferol-(diacetyl coumarylrhamnoside)with m/z 663.2 accumulated in the mesophyll layers (third and fourth) with uniform lateral distributions.

6674 Analytical Chemistry, Vol. 81, No. 16, August 15, 2009

Dow

nloa

ded

by G

EO

RG

E W

ASH

ING

TO

N U

NIV

on

Aug

ust 1

8, 2

009

Publ

ishe

d on

Jul

y 2,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

ac90

0745

e

CONCLUSIONSThis work reports the first example of 3D imaging MS at

atmospheric pressure and provides a method to correlate molecularcomposition with tissue morphology. This novel development was anontrivial combination of lateral imaging and molecular depthprofiling capabilities that had been independently presented in aprevious publication.21 Successful 3D imaging put dramatically higherstability requirements on the LAESI ion source. Data acquisitionprotocols and data reduction and analysis were also more complex.Finally, the amount and quality of information gathered in a 3Dexperiment is significantly higher. This is most obvious in the caseof the established spatial correlations of ion abundances.

LAESI 3D imaging MS revealed the distribution of secondarymetabolites in correlation with tissue specificity in plants.34 Weanticipate that with LAESI eventually water-containing organs orwhole-body sections of plants and animals, as well as humantissues, can be analyzed in 3D under ambient conditions formultiple chemical species.

In principle, working at atmospheric pressure also enables thestudy of live specimens. Although locally destructive, laser ablationof small, e.g., 300 µm diameter, areas of a larger plant or animalcan leave the entire organism viable. Even the ablated cells arealive and nonmodified up to the moment of analysis, occurringon the millisecond time scale, so their chemical composition canreflect their condition and functioning at the time of the ablation.

Although 3D ambient imaging with LAESI has been provenfeasible in plant tissues, further improvements are needed inspatial resolution. For example, major variations in the watercontent or tensile strength of tissues can affect the lateral imagingand depth profiling performance of the method. If these propertieschange in the tissue, a preselected laser fluence might result invariations in the ablation depth. An automated feedback mecha-nism could correct for these effects by recording the depth ofablation for each laser pulse and adjusting the laser energy and/or wavelength to minimize depth variations. With the currentlateral and depth resolutions of ∼300 and ∼30 µm, respectively,LAESI offers low spatial resolving power in comparison to opticaland established vacuum MS imaging techniques. Reduced spot

sizes are expected from a focusing system with low sphericalaberration or from coupling the laser energy into the samplethrough a sharpened optical fiber. These two approaches promisethe metabolic analysis of single cells with dimensions down to ∼10µm, while maintaining good signal-to-noise ratios in the mass spectra.LAESI imaging with these improved lateral and depth resolutionscan enhance our understanding of the spatial organization of tissuesas well as cell-to-cell communication and molecular transport.

Instantaneous information on the chemistry of biological andmedical specimens without sample treatment can be useful in avariety of applications. Diverse applications in plant biology, inpharmacokinetic studies of absorption, distribution, metabolism, andexcretion (ADME) of drugs, endogenous metabolites, and xenobi-otics are also expected to benefit from LAESI 3D imaging MS.

ACKNOWLEDGMENTThe authors are grateful for the financial support of this work

by the U.S. National Science Foundation under Grant No. 0719232,by the U.S. Department of Energy (DEFG02-01ER15129), and bythe W. M. Keck Foundation (041904).

SUPPORTING INFORMATION AVAILABLEAnimated GIF image of the microscope investigation of the S.

lynise tissue following LAESI sampling with a single laser pulsedelivered on the adaxial (upper) epidermis; differential interfer-ence contrast microscope images captured in the Z-stack modefor a series of focal planes with a 5 µm depth step to opticallymap the resulting ablation crater; these depth-resolved imageswere then combined into a GIF animation; the ablation depth wasdetermined to be ∼35 µm; an example for a broad Pearsonproduct-momentum cross-correlation analysis for the 3D molec-ular distributions of 26 ions (Table S1); mass spectra thatdemonstrate varying molecular composition as the leaf tissue ofS. lynise is ablated by six consecutive laser pulses (Figure S1);groups of metabolites with similar 3D distribution patters in theleaf of A. squarrosa (Figures S2-S4). This material is availablefree of charge via the Internet at http://pubs.acs.org.

Received for review April 7, 2009. Accepted June 10,2009.

AC900745E

(33) Schroder, G.; Wehinger, E.; Lukacin, R.; Wellmann, F.; Seefelder, W.;Schwab, W.; Schroder, J. Phytochemistry 2004, 65, 1085–1094.

(34) DellaPenna, D.; Last, R. L. Science 2008, 320, 479–481.

Table 2. Pearson Product-Moment Correlation Coefficients between the 3D Spatial Distributions of Ion Intensitiesfor 12 Selected m/z in an A. squarrosa Leaf

m/z 163.0a 285.1b 287.1b 301.1b 317.1b 325.1a 493.1b 501.1c 647.2a 663.2c 813.5 893.5a

163.0 1.00 0.16 0.11 0.11 0.17 0.87 0.14 0.60 0.73 0.38 0.50 0.51285.1 1.00 0.65 0.67 0.59 0.17 0.32 0.35 0.15 0.22 0.36 0.23287.1 1.00 0.94 0.88 0.11 0.49 0.26 0.14 0.15 0.14 0.08301.1 1.00 0.88 0.13 0.48 0.30 0.16 0.19 0.18 0.10317.1 1.00 0.19 0.64 0.35 0.22 0.21 0.18 0.14325.1 1.00 0.17 0.73 0.87 0.49 0.20 0.61493.1 1.00 0.32 0.20 0.16 0.16 0.16501.1 1.00 0.83 0.83 0.71 0.61647.2 1.00 0.71 0.57 0.55663.2 1.00 0.51 0.43813.5 1.00 0.65893.5 1.00

a See Figure S4 of the Supporting Information for the related distributions. b See Figure S2 of the Supporting Information for the relateddistributions. c See Figure S3 of the Supporting Information for the related distributions.

6675Analytical Chemistry, Vol. 81, No. 16, August 15, 2009

Dow

nloa

ded

by G

EO

RG

E W

ASH

ING

TO

N U

NIV

on

Aug

ust 1

8, 2

009

Publ

ishe

d on

Jul

y 2,

200

9 on

http

://pu

bs.a

cs.o

rg |

doi:

10.1

021/

ac90

0745

e